12 ieee transactions on engineering management, vol. 54, no. 1

12 IEEE TRANSACTIONS ON ENGINEERING MANAGEMENT, VOL. 54, NO. 1, FEBRUARY 2007

Selecting a Customization Strategy UnderCompetition: Mass Customization, Targeted Mass

Customization, and Product ProliferationHasan Cavusoglu, Huseyin Cavusoglu, and Srinivasan Raghunathan

Abstract—Customization requires not only an implementationof proper manufacturing systems but also a proper strategyregarding when firms should offer customized products andwhat the nature of customization should be. This paper questions1) whether customization is better than no customization, and,if so, 2) what kind of customization strategy firms should adoptunder competition. We find that customization is not optimalwhen the cost of soliciting customer preference information issufficiently high. When competing firms choose to customize,we show that firms target only certain customer segments withcustomized products. We also find that the optimal customizationstrategy may require firms to offer only a few discrete productvarieties. Despite the concern that customization may initiateprice wars because customization reduces product differentiation,we find that customization does not escalate the price competition,because aggressive price competition exacerbates cannibalization.Although customers within the product line of a firm are chargedhigher prices, we show that on average customers are better offwhen firms adopt customization. However, unless the customiza-tion is quite cheap, when firms choose to customize, we find thatfirms cannot generate more profits than when firms offer only asingle product.

Index Terms—Cannibalization, competition, customization, dif-ferentiation, flexible manufacturing systems, mass customization,operational flexibility cost, price competition.

I. INTRODUCTION

MASS production enables manufacturing of standardproducts in large quantities at low unit costs. Cus-

tomization, on the other hand, enables firms to provide goodsand services that match individual customers’ needs. Cus-tomization is often achieved by changing, assembling, ormodifying standard products according to customers’ desires[48]. Conventional manufacturing technologies typically serveone of these two production choices [19]. Dedicated machinesin assembly lines achieve mass production of a standardproduct with high efficiency (or low marginal production cost)while inhibiting flexibility, whereas flexible machines in job

Manuscript received September 1, 2004; revised April 1, 2005 and May 1,2005. Review of this manuscript was arranged by Special Issue DepartmentEditor, T. Blecker.

H. Cavusoglu is with the Sauder School of Business, The University of BritishColumbia Vancouver, BC V6T 1Z2, Canada (e-mail: [email protected]).

H. Cavusoglu and S. Raghunathan are with the School of Management,The University of Texas at Dallas, Richardson, TX 75083 USA (e-mail:[email protected]; [email protected]).

Digital Object Identifier 10.1109/TEM.2006.889064

shops facilitate production of different varieties of a productin batches while sacrificing efficiency. The advances in manu-facturing technologies, such as flexible manufacturing systems(FMS), computer-aided design/manufacturing (CAD/CAM),and just-in-time (JIT), have changed the ways organizationsproduce and market consumer goods. Manufacturers no longerneed to trade the efficiency of mass production for the flexibilityof customization. These new technologies have improved themanufacturer’s ability to produce customized products withouta significant increase in marginal production cost.

These new dynamics of manufacturing have enabled firmsto offer a large number of varieties in almost every conceiv-able product category. A study by Federal Reserve Bank ofDallas [14] reports that since the 1970s, the number of productvarieties has increased dramatically: Frito-Lay chips from 10to 78, personal computer models from 0 to 400, car modelsfrom 140 to 260, and car styles from 654 to 1212. The samestudy found that the average number of varieties offered bya single firm increased as well. Many firms offer multiplevarieties tailored to meet diverse customer tastes, partly totarget market segments unserved or poorly served by existingproducts. Customization also allows firms to charge higherprices for their offerings. Customers are frequently willingto pay a premium that reflects the added value of satisfactionthat arises from individualized solutions [10]. Although firmstake advantage of new capabilities of flexible manufacturingsystems, they follow different product variety (customization)strategies. While some firms offer two or more discrete productvarieties to serve their customer bases (i.e., product prolifera-tion), others offer every possible variant of the product withintheir customization scope (i.e., mass customization) [49].1 Forexample, a firm adopting the product proliferation strategy mayoffer a product in two colors- say, yellow and red- but a firmadopting the mass customization strategy may customize theproduct to any shade of color within the yellow-red band. Inspite of advances in manufacturing technologies that enablevarious forms of customization, it is not clear whether firmscan benefit from customization in the face of competition sincerival firms can also adopt similar customization strategies.Besides, increasing the number of product varieties can haveundesirable impacts on cost, quality, and responsiveness in amanufacturing environment [42], [58]. Therefore, firms shouldcarefully assess the cost and benefit of customization over no

1Both mass customization and product proliferation deliver goods and ser-vices that fit the needs of individual customers with a production efficiency thatalmost equals that of mass production [55].

0018-9391/$25.00 © 2006 IEEE

CAVUSOGLU et al.: SELECTING A CUSTOMIZATION STRATEGY UNDER COMPETITION 13

customization before investing in customization technologies.Another issue confronting firms today is the design of an ef-fective and profitable customization strategy. There are severalcustomization strategies from which firms can choose. Theselection of an appropriate customization strategy becomeseven more critical in today’s competitive markets becauseeach customization strategy can affect competition differently.Therefore, a firm designing its customization strategy mustanswer several questions, including the following:

1) Should the firm choose product proliferation, mass cus-tomization, or a hybrid strategy? Or is it better for the firmnot to customize at all?

2) How should the products be positioned within the firm’sproduct line?

3) What should be the price of each product variety offered?These decisions, of course, are influenced by both operational

factors and market characteristics. Operational factors includethe cost associated with investment in customization technolo-gies that provide flexible manufacturing capabilities and/or thecost of obtaining customer preferences before manufacturingcustomized products, depending on the firm’s choice of cus-tomization level. Market factors include competition, the extentto which a firm’s decisions affect competing firms; segmenta-tion, the extent to which a product variety extracts surplus fromcustomer(s) to whom it is targeted; and cannibalization, the ex-tent to which a product variety reduces a firm’s profits fromthe other varieties it produces. This paper integrates these fac-tors into a model for designing a firm’s optimal customizationstrategy.

As mentioned above, moving from a single variety tocustomization requires a firm to acquire unique operationalcharacteristics [58]. Firms have to be flexible enough to man-ufacture and distribute different varieties without significantlyincreasing the cost [58]. Depending on the requirementsof the products, firms may gain flexibility through variousmanufacturing techniques, such as modular design, platformcommonality, postponement, and configuration and/or lo-gistics techniques to achieve integration, coordination, andtransparency in the supply chain. Although these techniqueshelp firms increase efficiency in customization, introducingflexibility requires a considerable initial investment that wecall operational flexibility cost [28]. Production economics issuch that firms incur an increasing marginal cost to acquireadditional flexibility in manufacturing and logistics to offer alarger set of products that are tailored to heterogeneous con-sumer tastes [3]. Serving consumers with customized productsfirst requires identifying their unique needs [48]. In addition tooperational flexibility cost, firms incur information elicitationcost, which is the cost of interacting with customers to obtainspecific information about their preference structure [58]. Thiscost increases at the same marginal rate with the number ofcustomers for whom customized products are offered, becausegathering and processing customer information is similar foreach additional customer, irrespective of the collection methodused [20].

Despite the promise of customization to produce anythingthat a customer wants, diversity in customer tastes, in additionto customization costs, hinders offering customized products to

every customer. In reality, firms often customize according toa limited set of attributes along which customers’ preferencesvary. Following the convention in customization literature, weanalyze firms’ customization strategies by assuming that prod-ucts are customized based on an attribute, which is either asingle attribute or an aggregate attribute weighted over multipleattributes. Customization literature often uses a linear [32] ora circular [52] model of unit length to represent products thatdiffer on the customization attribute. A point in the line or inthe circle is considered to be a product variety. Customer pref-erences, or their ideal products, are assumed to be uniformlydistributed along the unit line or circle. In this model, no cus-tomization strategy is represented as a point corresponding toa single product offered by the firm, whereas product prolif-eration strategy is captured by a set of discrete points corre-sponding to the product varieties offered by the firm. Mass cus-tomization strategy is represented by one continuous segmentalong the line or the circle. The firm adopting mass customiza-tion offers every product variety within this segment, known asthe customization scope of the firm [46]. Though customiza-tion for the entire product space may be impossible to achieve,a firm may be able to achieve customization in more than onesegment of the product space. That is, a firm may have multiplecustomization scopes, each representing a range of customizedproducts, instead of product varieties in one continuous cus-tomization segment. We denote this hybrid strategy as targetedmass customization. Analysis of targeted mass customizationis new to the customization literature and may explain why, insome markets, firms produce ranges of customized product va-rieties whereas in other markets firms only produce a number ofdiscrete product varieties when they customize.

Our analysis of customization strategies under competitionoffers several insights into when firms should offer customizedproducts, what the nature of customization should be, andhow the selected strategy affects product line design, pricecompetition, rent extraction, and social welfare. We find thatthe availability of flexible manufacturing systems is not a suffi-cient reason for firms to adopt customization. Customization issuboptimal when the cost of soliciting customer preference in-formation is very high. Contrary to popular belief that firms willextend their product offerings as customization gets cheaper,we find that firms do not increase their customization scopesbeyond a certain level even when the cost of customization doesdecrease. While consumers always benefit from customization,we find that, unless the cost of customization is sufficientlylow, firms are worse off when they offer customized productsinstead of a single product. Competition between firms is thereason firms end up with such an inferior outcome even iffirms adopt the optimal level of customization. When firmscan aim customization at specific consumer segments, weshow that offering a single customization scope may not beoptimal. Firms offer either several customization scopes orseveral discrete products, depending on the cost of obtainingcustomer preferences vis-à-vis the price premiums associatedwith customized products. Compared to mass customization,targeted mass customization allows firms to lock in somecustomers between their customization scopes. This, in turn,enables firms to enjoy local monopolies within their product


lines. Firms also save on customization cost by not offeringcustomized product varieties to those customers. We find thatfirms always accommodate some noncustomization segments-those who are served with standard product varietiesin theirproduct lines so long as customization is not absolutely free.Customization reduces differentiation between firms’ productlines. However, this does not exacerbate the price competition.

Our paper contributes to the extant literature in several ways.First, we show that customization is not always the optimalchoice for competing firms. Second, we find that firms may bebetter off by focusing on several discontinuous segments of themarket instead of one continuous segment when customizing.Third, we show that the optimal customization strategy may beto introduce several product varieties instead of a range of prod-ucts within a segment or multiple segments.

In Section II and III, we first summarize the relevant litera-ture. Then, we describe our model and analyze the first modelthat compares customization with no customization to answerwhether customization is better than none. Next, to find thebest customization strategy, we analyze a second model thatcaptures several customization strategies as special cases. Afterthat, we compare customization strategies along several dimen-sions, such as price competition, rent extraction, and consumerwelfare. We end the paper with some concluding remarks, man-agerial implications, limitations, and future research directions.

II. LITERATURE REVIEW

Our work on customization is related to the product linedesign studies that examine the breadth and positioning ofproduct lines of competing firms. Since product line designstrategies affect firms’ decisions on production and mar-keting-mix elements, such as pricing and distribution channels,management of product variety has received great emphasis inthe marketing, economics, and operations literature [31], [34].The following paragraphs summarize the existing literaturefrom these streams.

Because the idea of product variety conflicted with theconcept of mass production, the main focus in the early daysof manufacturing was on producing a single product. Cor-respondingly, much of the economics literature investigatedmonopolistic competition in which each firm offers one varietyin the product space [21], [36], [57]. As competition intensifiedin many industries, and manufacturing technologies made itpossible to increase variety without losing efficiency, the mar-keting and economics literature started to focus on the effect ofproduct variety strategies on competition and how firms shouldcope with the adverse effects of competition by managing theircustomization strategies. Lancaster [37] reviewed possibleproduct variety strategies. Brander and Eaton [7] analyzedthe competition in product positioning between multiproductfirms by assuming that each firm in a duopoly produces twoproducts. They showed that firms could tailor products for dif-ferent market segments and accordingly charge higher prices.However, Martinez-Giralt and Neven [44] showed that, whenfaced with the choice of offering either one or two products,firms would choose one product because price competitionoutweighs the benefits of market segmentation when each firm

offers two products. Recently, Cavusoglu and Raghunathan[9] concluded that the result in [44] is mainly driven by thespecific functional form chosen for transportation cost and thatproduct proliferation in fact can be the optimal strategy. Theabove studies focus on the customization strategies wherebyfirms can produce only distinct product varieties in the productspace. Another stream of research investigates customizationstrategies in which firms offer all possible varieties within theircustomization scopes [5], [54]. Beckman [5] and Thisse andVives [54] analyze the strategy by which firms customize byredesigning a basic product to suit consumer tastes. Since,in this case, customization is achieved through reconfiguringfinished products, their models do not consider any ramificationof customization in manufacturing environments. The extentof product customization is contingent upon a firm’s flexibilityto reconfigure its manufacturing resources, not its ability to re-configure the finished products. This manufacturing flexibilitycan be measured as the range of product varieties offered bythe firms. Mussa and Rosen [46] represent customization as acontinuous spectrum of product characteristics. Although theirmodel allows the firm to produce a range of product varieties,it ignores the cost of customization. In a later study, Dewan etal. [20] incorporate the missing cost of customization and an-alyze single-scope product customization. However, the singlecustomization region that was exogenously imposed in [20],[46] does not allow firms to customize in some niche segmentsof the market. In this paper, unlike early studies, we relax thesingle-scope assumption and allow firms to focus on multipleniche scopes. This setup also enables firms to adopt productproliferation, which is to produce multiple discrete products.Our multiscope customization framework allows firms to tradeoff the cost of improving flexibility of production against thesetup cost to accommodate multiscope customization. In fact,we show that firms may be better off by adopting targeted masscustomization.

In marketing, product line design studies have focused on thenumber of products to be introduced, their positions, and theirprices [27], [33], [37]. The customers were assumed to be dif-ferentiated either vertically or horizontally based on product at-tributes [2], [8], [12], [18], [29], [45], [56]. Customers are ver-tically differentiated if all customers agree that the attribute ofdifferentiation is the quality, and that customers differ in theirwillingness to pay more for higher quality. However, customersare horizontally differentiated if the differentiation occurs basedon product attributes other than product quality. For example,customers are said to be horizontally differentiated if their pref-erences change with respect to the size of the back-pockets ofa pair of jeans. In other words, the size of the pocket is not re-garded as a determinant of quality. Following the seminal workof Hotelling [32], many studies in the product variety litera-ture have considered horizontal differentiation. In this paper,we also characterize customization in a horizontally differen-tiated market. Prior studies have generally assumed customerself-selection [45]. That is, each customer is entitled to choosethe product variety that gives the highest utility. Since a cus-tomer can choose to purchase a product intended for a differentcustomer segment if he/she gets a higher utility from it, canni-balization- competition between product varieties offered by the


same firm- can pose challenges to firms offering multiple vari-eties. Hence, because customization requires careful pricing andlocation decisions, when we analyze customization strategiesamong competing firms, we explicitly consider cannibalizationand eliminate it by using an appropriate pricing scheme.

Since customers can have diverse valuations for the sameproduct, offering a standard product with a customized price hasrecently been discussed as an effective way to sustain competi-tiveness in some markets. Today, firms such as direct marketingfirms, catalog retailers, credit card companies, and long-dis-tance phone companies, can determine how much each cus-tomer is willing to pay and, in response, offer their products atan individual price that extracts the highest rent from each cus-tomer [11]. However, such a strategy only offers a customizedprice without a customized product. In this paper, firms notonly produce customized varieties but also charge customizedprices. Therefore, because customization strategies are produc-tion driven, they take into account both cost elements and pricepremiums associated with customization.

Much of the work in the operations management area hasfocused on manufacturing strategies (techniques and methods)that support a high product variety [22], [38], [40], [41].Krishnan and Ulrich [35] review product development anddesign literature. Lee [39] proposes postponement as a strategyto alleviate prohibitive costs of producing large product vari-eties. Desai et al. [17] propose a modular design that enablessharing components to maintain quality. Netessine and Rudi[47] examine the tradeoff between production and supply chaincosts in decisions on product variety. Smith and Agrawal [53]consider joint inventory and product line design. In this paper,instead of analyzing a specific technique or method to supportcustomization, we incorporate the cost of customization intoour model regardless of techniques used to support customiza-tion. This level of abstraction allows us to simultaneouslycapture marketing and manufacturing aspects of customizationto determine the optimal customization strategy.

Empirical studies on product variety aim mainly at deter-mining external and internal factors that influence firms’ productvariety strategies. In one group of studies, market forces and in-dustry structure are used to explain firms’ decisions on productvariety. Connor [13] concludes that imperfect market structureslead to higher levels of product proliferation in many food cat-egories. Bayus and Putsis [4] find that the extent of productvariety in the computer industry depends on the firm’s marketshare and the costs of supplies. In a later paper, Putsis and Bayus[50] find that firms increase product varieties when industry bar-riers are low and market opportunities are perceived to exist. Inthe other group of studies, the focus is on identifying and mea-suring the variety-related product losses. Fisher et al. [24] ob-serve that greater product variety causes an increase in overheadcost, quality problems, the risk of stockouts, and the need for re-work in the automobile industry. They suggest that firms needto commit more resources to support smooth operations as theyincrease the number of product varieties. MacDuffie et al. [43]discover that labor productivity decreases as the mean optioncontent per vehicle increases. Fisher and Ittner [23] find that thegreater day-to-day variability in option content, but not the meanoption content per car, significantly increases total labor hours

per car, overhead hour per car, assembly line downtime, and in-ventory level. They also show that bundling options improvesperformance as it reduces the random variability in options in-cluded in the cars manufactured. Randall and Ulrich [51], usingdata from the U.S. bicycle industry, show that a firm’s productvariety strategy depends on its supply chain cost, which com-prises production cost and cost of reaching the target marketsuccessfully. However, they find no evidence to suggest that of-fering multiple varieties through strategies such as customiza-tion or variety postponement results in higher firm performance.Another study also finds that increasing variety using platformapproach does not always increase the profitability of productline [30]. Given our finding on profitability of customization,we might speculate that firms in those industries did incur highcustomization costs that prevented them from enjoying the ben-efits of customization, compared to the single product case.

III. MODEL

We consider two symmetric firms that are determining theircustomization and pricing strategies. We model the productspace, which represents all possible variations of a specificproduct, and uniform distribution of customer preferences,2along a unit circle. The circular model of product space was firstintroduced by Salop [52] to eliminate the end point difficultiesof Hotelling’s linear model [32].3 In our circular model, firmsare located at the opposite ends. Each firm can offer one or moreproducts within its product space. We model the competitionbetween firms as a two-stage game. In the first stage, firmssimultaneously choose their customization strategies. Theysimultaneously determine their prices in the second stage. InSection III, we first analyze whether customization is preferredover no customization when customization technologies arereadily available to both firms. If customization is an optimalstrategy when there is competition, we next address whatkind of customization strategy—namely, mass customization,targeted mass customization, or product proliferation—firmsshould adopt. Finally we compare various customizationstrategies to analyze their effects on product line design, pricecompetition, rent extraction, and consumer welfare.

A. Should a Firm Customize at All?: Customization VersusNo Customization

In this section we capture customization within a singlescope. When a firm chooses to customize, it offers every cus-tomer within its customization scope (customization segment)a product that perfectly fits his/her preference. We capturethe customization scope of a firm as a continuous arc alongthe circle that represents its product space. The length of cus-tomization scope determines the cost of customization. For acustomization scope of length , firms incur a cost of .4The first term is associated with operational flexibility that en-compasses manufacturing, distribution, and customer service

2This is a standard assumption in product variety literature in economics. Al-though some operations literature uses a beta distribution to represent customerpreference, it has been used only in a monopolistic setting.

3Our results also apply to the linear model of product space.4There may also be a fixed production cost that is independent of x. However,

this can be normalized to zero.


Fig. 1. Conventional and Customization Segments in Duopoly.

costs [28]. The second term models the information elicita-tion cost of obtaining customer preferences [58]. A customerlocated distance away from the location of a product, whichis priced at , receives a utility of , where is thereservation price that a customer is willing to pay and is themisfit cost per unit distance.5 Each customer in the market whohas a unitary demand chooses the product that gives him/herthe highest utility. A customer located in a firm’s customizationscope can purchase a product variety that perfectly fits his/herpreference, but a customer who is located outside the firm’scustomization scope cannot. Customer segments served by thefirm but not covered by the customization scope are often de-noted as conventional segments, as depicted in Fig. 1. Note thatno customization is a special case of the customization modeldepicted in this section. If the firm’s optimal customizationscope is zero, it implies that the firm prefers no customizationover customization, because customization scope of zero lengtheffectively means that the firm offers only a single productvariety.

Firm charges for standard products, which are theproducts positioned at the edges of its customization scope.It charges for the product located (where )units of distance away from the standard product within thecustomization scope. A firm’s profit expression under masscustomization can be specified as

(1)

5Misfit cost captures how sensitive the customer is to product differences.Through customization, firms charge more for customized products than stan-dard products. As we show later in the paper, misfit cost corresponds to pricepremium associated with customized products.

Intuitively, a firm can charge higher prices for product vari-eties within its customization scope than for its standard prod-ucts. We can show that the firm chooses a linear discriminatorypricing scheme for the product varieties withinits customization scope. This pricing scheme follows from theobservation that firm can extract as much as the disutility thata customer would suffer if he/she were to purchase a standardproduct.6 The term represents the price premium associatedwith customization, because a customer who is located dis-tance away from the standard product is willing to pay this muchmore to get his/her ideal product.

In the first stage, firms choose the sizes of their customizationscopes simultaneously by maximizing their respective profits.

(2)

In the second stage, they determine prices for their productssimultaneously

(3)

We solve the problem using backward induction. Solving op-timization problems given in (3) simultaneously with respect toprices leads to the pricing of standard products as

(4)

Before moving to the first stage, we need to ensure thatthe pricing strategy profile in (4) is subgame perfect. For astrategy profile in a multistage game to be subgame perfect, noplayer should gain by deviating from the action prescribed inthe strategy profile at any stage of the game (Fudenberg andTirole [25, p. 108]). The price expressions in (4) characterizethe pure strategy equilibrium in the price subgame if and onlyif neither firm can be better off by charging a different price.Since the customization problem is a two-stage game, to verifysubgame perfectness, we need to check whether any choiceof customization scope sizes in the first stage would lead oneof the competing firms to deviate from the stated equilibriumpricing strategy. D’Aspremont and Gabszewicz [16] show, inthe context of single product case, that if the firms’ productsare closely located, then there is no pure strategy equilibriumin the price subgame. In our single scope customization model,we need to check if attaining equilibrium in the price subgame

6If firm i charges higher than the price shown in the linear pricing scheme,the customer will choose to purchase the standard product. If it charges a lowerprice, the firm will not extract the whole rent. Therefore, the linear discrimina-tory pricing scheme is optimal.

if

ifif

(5)


ifotherwise

(6)

can be an issue. Formally, given the customization strategy, forto be a pure strategy equilibrium for firm , we must have

, for any . The profit expression givenin (1) assumes that the subgame perfection is not violated.By considering deviations in the pricing stage, the completecharacterization of the profit for firm can be written in termsof firms’ scope sizes and prices as shown in (5) at the bottomof the previous page.

It is clear that the profit function exhibits two discontinuities.Therefore, the optimal price reaction can be expressed as shownin (6) at the top of the page.

If , the customers in the marketsegment that firm is supposed to serve would prefer to buyfrom firm . Hence, firm captures the entire market. The priceexpression in (4) is the equilibrium if and only if firm ’s profitunder these prices is not less than the profit when it capturesthe entire market by deviating from the price equilibrium. Fora given customization scope, the profit of firm under the priceequilibrium is

If firm chooses to chargewhen the firm chooses the optimum price of

, the profit of firm for a givencustomization scope is

A quick comparison of and reveals that any deviationfrom the pricing equilibrium for firm is not profitable (i.e.,

) under the following condition:

(7)

A similar condition can be obtained for firm as

(8)

Thus, when either (7) or (8) is not satisfied, then at least onefirm has an incentive to deviate from the pricing strategy given

in (4). It is clear that the price equilibrium is stable when condi-tions (7) and (8) are satisfied. Further, firms have to ensure thatcustomization scopes are nonnegative. In other words

(9)

When (7)–(9) are satisfied, the pricing game is subgame per-fect. Thus, to ensure the stability of pure strategy price equilib-rium, firms should solve the following constrained optimizationproblem simultaneously

(10)

The optimal solution to the single scope customizationproblem is given in Proposition 1.

Proposition 1: The two-stage game in customization scopesand conventional market prices has a unique subgame perfectequilibrium in pure strategies of

if &

if &

if

(11)

Proof: Refer to the Appendix.From (11), the equilibrium profit of firm can be found as

if &

if &if

(12)

Fig. 2 illustrates Proposition 1. As can be seen from the figure,firms’ customization cost ( and ) relative to customers’ misfitcost determines whether firms choose customization overno customization, and if they do, the extent of customization.Proposition 1 offers several insights. First, customization isoptimal only when gathering customer preferences is relativelycheap compared to customers’ marginal disutility associatedwith buying a nonideal product. That is, if it is too costly toelicit customer preference information, it is worthwhile not toadopt customization. In such a case, each firm is better off byoffering a standard product. However, Internet technologies,such as online registration, cookies, and filtering tools, enablefirms to understand customer preferences at low costs ( canbe actually very small). Therefore, we can say that if firms usethese technologies effectively, they will find customization areality in many product categories, irrespective of how expen-sive it is to produce customized products. The cost of flexiblemanufacturing will only affect the extent of customization.Second, we show that the common belief that “with moreadvanced customization technologies available, the firms willexpand their customization scopes” [20] might not be always


Fig. 2. Equilibrium Regions with Respect to Cost Parameters.

true. Customization scope of firms does not keep increasingas decreases. In fact, once the customization scope reaches1/4, a further decrease in does not change the customizationscope. Our results show that with customization in duopolisticcompetition, a maximum of fifty percent of customers areoffered customized products and the other fifty percent buystandard products, irrespective of the cost of customization.7Third, when firms prefer customization (i.e., ), a declinein the cost of customization (the cost of operational flexibilityor the cost of eliciting customer information) does not alwayslead to more competition between firms and lower profits forfirms. Although a reduction in or first leads to a decreasein the firms’ profits, we find that the decrease in the firms’profits may not be indefinite. Indeed, once the customizationscope reaches its maximum limit (because of a drop in oror both), a further decrease in flexibility cost and/or elicitationcost always results in an increase in profits. This is becausea decrease in the cost of customization increases competitionbetween firms and results in more customization. However,since firms do not extend their customization scopes indefi-nitely, after reaching the maximum customization level, firmsstart enjoying the positive effect of cost savings on their bottomlines. The above results are the direct interpretation of partialderivatives of a firm’s profit and customization scope size withrespect to customization cost parameters ( and ).

One can measure the value of customization over no cus-tomization in terms of additional profit that it generates forfirms. The value of customization can be obtained by com-paring profits of firms when they adopt customization withprofits when they offer only a single product. The followingproposition summarizes this comparison.

Proposition 2: Relative to the case in which both firms offera single product, firms are 1) always worse off when they adopt

7In this paper, we assume that operational flexibility cost is a quadraticfunction of the customization scope. This form implies that marginal costof customization increases as the scope of customization increases (i.e.,@ (operational exibility cost)=@x > 0). However, any functionalform with increasing marginal cost, such as e , is sufficient to get the samequalitative results. The function form of operational flexibility cost only affectsthe value of the maximum customization scope.

customization with less than the maximum customization scope,and 2) better off when they adopt customization with the max-imum customization scope if and only if firms can recover thecost of customization through price premiums (i.e.,

).This result does not imply that firms do not adopt customiza-

tion unless it is more profitable than no customization. Cus-tomization is the outcome of simultaneous decisions made bythe firms under competition. Competition leads firms to cus-tomize when . It is easy to show that each firm prefersto customize if its rival customizes. Proposition 2 implies thatfirms are not always better off with customization. This unin-tuitive result can be explained by analyzing the forces that af-fect profit expressions of firms. When firms customize with ascope of size less than 1/4, although they get additional rentfrom customers who purchase customized products, the addi-tional revenue through price premiums is not enough to coverthe cost of customization. The additional revenue required tooffset the cost of customization is generated only when firmscustomize at the maximum extent and the cost of customizationis sufficiently small, as specified in Proposition 2. Hence, cus-tomization in the duopoly case can lead competing firms to thePrisoner’s dilemma.8

In this section we assume that each firm offers a singlecustomization scope. However, offering a single customizationscope, even if it turns out to be better than no customization,may not be the optimal customization strategy. The optimalcustomization strategy may require firms to produce multiplediscrete products or to customize with multiple customizationscopes. In Section III-B we formulate and solve the customiza-tion problem with a more general model that endogenizesthe number of customization scopes to accommodate variouscustomization strategies.

B. Selecting the Optimal Customization Strategy

In this model, we assume that firm introduces cus-tomization scopes on each semi-circle,9 as shown in Fig. 3.Each customization scope is represented with an arc in the unitcircle. The size of customization scope of firm is denoted as

.10 As in Section III-A, the products at the boundary of eachcustomization scope are referred to as conventional products.Between the conventional products in each customizationscope, firms produce customized products. Competing firmslocate their closest conventional product varieties distanceaway from each other. Firm leaves distancebetween its th and th customization scopes. Customersin these segments are not offered customized products.

8Prisoner’s Dilemma is a famous game in game theory to explain a conflictbetween individual and group rationality and to show how a group whose mem-bers pursue rational self-interest may all end up worse off than a group whosemembers act contrary to rational self-interest. The game gets its name froma hypothetical situation where two criminals arrested under the suspicion ofhaving committed a crime together. Each prisoner has only two choices: co-operate (i.e., not confessing to the police) or defect (i.e., confessing to the po-lice). Since neither prisoner knows what the other prisoner would choose, eveneach prisoner gets more jail time when both defect than that when neither de-fects, each prisoner prefers to defect in the equilibrium. The payoff structure ofcustomization in duopoly resembles the classical prisoner’s dilemma game insome cases. Since the profit when neither firm customizes is less than the profitof the customizing firm whose competitor chooses not to customize, each firmis “tempted” to customize in those cases. Hence, in the equilibrium, both firmscustomize and end up with less profit than the profit they would have gained hadboth chosen not to customize.


Fig. 3. Customization Structure.

Firm charges and for its conventional products lo-cated at the boundaries of its customization scope . The priceof a customized product located at a distance from the con-ventional product priced at is given by . Similar tothe setup in Section III-A, firm incurs a cost offor offering customization scope . Compared to a manufac-turing environment that enables customization with the singlescope of size , a manufacturing environment enabling cus-tomization with multiple scopes of total size is less flexible.This is because a manufacturing plant achieving multiple-scopecustomization cannot readily move from producing product va-rieties in one customization scope to producing product vari-eties of the others. To do so, the firm has to incur a setup costto accommodate multiscope customization in a semi-flexibleenvironment, such as cost associated with model changeover,changing fixtures, or underutilization of machinery [1, p. 58].For example, as the number of body styles increases within anautomobile plant, higher setup costs are incurred at the bodyframing area due to switching between body styles [43, p. 353].We assume that firm incurs a setup cost of . We do notimpose any functional form for . Our results work with any

that increases in . Then the profit of firm can bewritten as11

9Because of the symmetric arrangements of products on each half of thecircle, firm j has a total of 2N customization scopes.

10Index i starts with the scope closest to competitor’s product line and in-creases as we move to the center of the firm’s product line.

11� represents the profit generated from the half of firm j’s market segmentand j 6= �j 2 f1; 2g.

(13)

This customization model incorporates various customiza-tion strategies speculated in previous sections as special cases.For instance, when the optimal customization strategy requiresa firm to place its customization scopes adjacent to each otherwithout any noncustomization segment between them (i.e.,

and ), this modelreduces to the single-scope (mass) customization. If the place-ment of customization scopes is such that there is at least onenoncustomization segment between adjacent customizationscopes (i.e., and ), thenmultiple-scope (targeted mass) customization is the optimalcustomization strategy. Finally, when the optimal customiza-tion strategy leads to multiple discrete product varieties (i.e.,

), then the model reduces to product proliferation.Hence, these three customization strategies- mass customiza-tion, targeted mass customization, and product proliferation-collectively cover all possibilities.

The game is played in two stages. In the first stage, firmssimultaneously determine the numbers, locations, and sizes ofcustomization scopes to maximize their profits

(14)

In the second stage, they simultaneously choose the prices oftheir products

(15)

We first solve the pricing problem given in (15) and then thecustomization problem given in (14). The pricing problem forthe multiple-scope customization is similar to that for the single-scope customization. Therefore, we just present the solution tothe pricing problem. To extract the maximum rent possible, firm

sets the price of the conventional product in customizationscope using the following scheme:

(16)

Similarly, the price of a customized product located at dis-tance away from a conventional product in scope of firmis set at

(17)

Therefore, the conventional product at the other end in scopeof firm (i.e., the one located further from the competitor) is

priced at

(18)


When we substitute (16)–(18) into (13) and solve the problemgiven in (15), we find the optimal prices as

(19)

where .In order to ensure stability of the price equilibrium, we have

to impose some conditions on . Following the steps describedearlier in Section III-A, we obtain the following constraints:

(20)

(21)

We also know that each scope has a nonnegative size. In otherwords,

(22)

In the first stage, firms solve the optimization problem in (14)such that (20)–(22) are satisfied. The solution to this problem issummarized in Proposition 3.

Proposition 3: Optimal customization strategy requires that1) each firm introduces number of scopes where

. 2) The location of each scope is . 3)

The sizes of the scopes are

if

if .(23)

We see from Proposition 3 that when , firms offeronly discrete products. Specifically, each firm offers con-ventional products with no customization scope around them.This result contradicts the belief that a firm always prefershaving customization with scopes [20]. When firms chooseto introduce standard products, they prefer to attain neithermaximum differentiation nor minimum differentiation, but acertain level of differentiation. This result is analogous withthe findings in [16]. When , each firm offerscustomization scopes. Each customization scope has the samelength, and is separated at an equal distance from the edges oftwo of the firm’s adjacent customization scopes.

Fig. 4 illustrates Proposition 4 for the case in which .We can see that when firms prefer customization with scopes,each firm introduces customization regions at the boundaries ofits product line and does not introduce any customization regionat the center of its product line. When firms prefer product pro-liferation, each firm introduces only two conventional productsat its boundaries.

Corollary 1: When demand is inelastic (i.e., is high) and/oreliciting customer preference information is not costly (i.e.,is low), firms choose to implement customization with multiplescopes. However, when demand is elastic and/or eliciting cus-tomer information is costly, firms choose to introduce a set ofconventional product varieties.

This corollary describes the relationship between operationaland market characteristics in choosing the optimal customiza-tion strategy. The demand elasticity, a market characteristic, orthe cost of eliciting customer information, an operational char-acteristic, can be enough to determine whether the firm is betteroff with customization with multiple scopes (mass or targetedmass) or product proliferation. If firms serve customers whoare not sensitive to price changes or have access to inexpen-sive technology to collect customer preference information, theyshould implement customization with multiple scopes. Other-wise, firms should adopt product proliferation as a customiza-tion strategy.

Customization in some industries confirms our finding inCorollary 1. In the breakfast cereal industry, although customertastes are heterogeneous, customers are not sensitive to changesin ingredients. For instance, although they may prefer wheatover corn, they may not choose cereal based on Vitamin B6content (i.e., is low). This argument is consistent with theempirical finding that consumers are sensitive to price changein cereals [26]. As a result, as predicted in our model, firmsproduce various breakfast cereal brands (varieties) for differentgrain types. Corollary 1 also explains why traditional footwearmanufacturers produce shoe pairs based only on foot lengthsince it is very costly to obtain exact fit characteristics ofeach customer (i.e., is high). Although 3-D foot scanningsystems can obtain these fit measures, it is prohibitively costlyto install such systems. If customers themselves knew their fitmeasures and revealed them to manufacturers, manufacturerscould adopt (mass or targeted mass) customization. Given thatmultiple-scope customization is optimal (i.e., is low or ishigh), the following corollary shows when a firm should prefermass customization over targeted mass customization.

Corollary 2: Unless operational flexibility cost and costof eliciting customer information are both zero (i.e.,and ), firms will always have customers between theircustomization segments who will be served with conventionalproducts.

This is because the size of each customization segment isequal to the distance between the edges of two adjacent seg-ments (i.e., ) only when both and are zero. Thisimplies that mass customization is the optimal customizationstrategy only when customization is costless. Otherwise (i.e.,when either or is not zero) each firm prefers targeted masscustomization over mass customization and offers customizedproduct varieties to frac-tion of customers. The remaining customers are served with theconventional products. Most firms offering customized prod-ucts today do not produce every possible variety that customersprefer to buy. Consider Factory 121, a Switch watch maker, asan example. The company allows a customer to design a watchfor his/her unique taste by choosing among available alterna-tives for each component such as case, bezel, crown, dial, andstrap. The combinations offered are almost infinite [48]. How-ever, not all combinations are allowed. Factory 121 eliminatesthe cost of eliciting customer information by using an interac-tive website. It also reduces operational flexibility cost consid-erably by starting the design process with a partially assem-bled watch. Since cost of customization is not exactly zero, ourmodel may explain why not all combinations are allowed and


Fig. 4. Locations of Products Along the Unit Circle When N = 1.

customers who prefer those combinations can only purchasethe variety which is closest to their ideal product variety. Theresult in Corollary 2 is very important since it implies that, al-though the mass customization can be better than no customiza-tion as described in Section III-A, mass customization cannot bebetter than targeted mass customization if customization is notcostless.

Corollary 2 provides important insights for customization ofdigital goods.12 Unlike traditional (nondigital) goods, once adigital product is ready, it is easy to modify it to meet the needsof various customer types. Therefore, in the case of customiza-tion of digital goods, is negligible. In addition, widespreadavailability of personalization services on the Internet makescollecting customer preferences almost costless. By creatingpersonal profiles, customers reveal their preferences to man-ufacturers. So can be zero for some digital goods13. Thus,we can conclude that mass customization is the best strategyfor customization of digital goods as long as eliciting prefer-ence structure of customers and delivering digital goods forindividual tastes are costless. For instance, many financialportals, such as Yahoo Finance, can be personalized based onconsumers’ preference to see stock prices, market announce-ments, performance ratings, and expert recommendationsregarding their portfolios. With the help of its sophisticatedweb technologies, Yahoo is able to deliver mass customizedcontents satisfying diverse needs of customers. Mass cus-tomization can also be the best strategy for some nondigitalgoods. Customers can design their own products by interactingwith online design sites. For instance, Reflect.com sells cus-tomized cosmetic products [15]. The bases for customizationof many cosmetic products are color and shininess. Because asystem that can produce red-shiny lipsticks is flexible enoughto produce pink-matte lipsticks, and because customers revealtheir preferences directly through the site, customization forReflect.com is almost costless. Therefore, Reflect.com adopts amass customization strategy as predicted by our model.

12We thank the reviewers for bringing this issue to our attention. In this paper,we use the term “digital goods” to refer to products and services whose contentis primarily information.

13Not every digital product can be mass customized. For example, the useof the Internet may not be enough to get customer preferences for complexsoftware without incurring any cost, such as operating system software.Therefore, as predicted by our model, operating system software vendors preferproduct proliferation, i.e., having multiple versions: home-user version andprofessional-user version, to mass customization.

C. Comparison of Customization Strategies

The difference between single-scope (mass) customization(as described in Section III-A) and multiple-scope (targetedmass) customization (as described in Section III-B) has todo with a firm’s ability to customize in certain ranges ofits product line. If operational capabilities enable a firm toapply its customization efforts to several specific niches, thefirm, in general, is strictly better off adopting multiple-scopecustomization instead of single-scope customization, sincesingle-scope customization is a special case of multiple-scopecustomization. In this section we specifically compare masscustomization and targeted mass customization. Our purposeis to see how the ability to target certain niches affects theextent of product line, price competition, rent extraction, andconsumer welfare.

Since firms are identical, it is not surprising that each firmserves half of the market, whether firms adopt mass customiza-tion or targeted mass customization. We know that the numberof customers that a firm serves with customized products isnot more than 1/4 of the total market in any customizationscheme. However, different from mass customization, a firm’sproduct line always covers 1/4 of the total market in targetedmass customization. The reason is that firms in targeted masscustomization can lock in some customers within their productline, thereby creating local monopolies and charging morethan what they would have charged if they had served thesecustomers in mass customization. Therefore, each firm hasa tendency to locate its customization segments (multiplescopes) in such a way that the noncustomization segmentsbetween them are maximized. Since competing firms cannotlocate their products closer than 1/4 of the attribute space, theproduct line always covers 1/4 of the market in the targetedmass customization. However, in mass customization, sincethere is only one customization segment (single scope), firmscannot have customer segments served with standard productswithin their customization scope. The decision to expand aproduct line is made based purely on the tradeoff between themarginal cost of expanding this single customization scope andthe marginal revenue that the firm gets by offering customizedproducts to additional customers. Hence, a firm’s product linein mass customization covers always less than or equal to 1/4of the total market.

Since a larger product line decreases differentiation betweencompeting products, one would expect that expanding the


Fig. 5. Pricing Under (i) Mass Customization and (ii) Targeted Mass Customization.

product line intensifies the price competition. However, weknow that the prices of the closest conventional products ofcompeting firms are in both mass customization and tar-geted mass customization. This is also the price if firms were tooffer only one standard product without engaging in any formof customization.

Corollary 3: Neither mass customization nor targeted masscustomization intensifies the price competition between thecompeting firms.

The intuition behind this result is as follows. Since firms cancharge higher prices to those customers who lie within theirproduct lines, they do not compete aggressively for the marginalcustomers who are located between their product lines. If firmsget into price competition, then the ensuing competition affectsnot only the prices of varieties produced by competing firms-those standard varieties that compete directly against each other-but also the prices of other customized varieties produced by thesame firm because of cannibalization problem. Overall, the firmhas more incentive to prevent cannibalization than to escalatethe price competition. Therefore, customization in both formsdoes not intensify the price competition.

Mass customization is also different from targeted masscustomization in terms of the amount of total rent extractedfrom customers. In mass customization, because the firmprovides product varieties in its product line that exactly fiteach customer’s taste, the firm can appropriate maximum rentfrom customers within its customization segment. However,the firm cannot extract the maximum rent from customers whoare outside of the firm’s product line since they only get con-ventional products. Similar to mass customization, in targetedmass customization, the firm can extract the maximum rentfrom customers who are provided customized products withinits product line. However, because some segments within thefirm’s product line are offered conventional products, targeted

mass customization also cannot appropriate the maximumrent from those customers. But, the rent that is extracted fromcustomers who get conventional products in targeted masscustomization is higher than the rent that is extracted fromcustomers who get conventional products in mass customiza-tion. Customers who purchase conventional products in masscustomization are charged due to competition from the rivalfirm, while customers who purchase conventional products intargeted mass customization are charged higher than dueto local monopolies. Hence, the firm extracts higher rent intargeted mass customization than in mass customization.

Corollary 4: Assuming the same extent of customizationscope, a firm adopted targeted mass customization extracts morerent than that a firm adopted (single-scope) mass customization.

Fig. 5 shows the locations and prices of product varietiesin mass and targeted mass customization, assuming that bothcustomization strategies have the same extent of customiza-tion segment (i.e., ). Note that the area under pricefigure represents the total rent extracted in each customizationstrategy. Hence, it is easy to see from Fig. 5 that a firm extractshigher total rent in targeted customization compared to masscustomization.

We know that customization may not result in any additionalprofit to competing firms. Does this finding imply that customersaccrue the benefit of customization? Irrespective of customiza-tion strategy, customers who are located between the productvarieties of competing firms enjoy additional surplus with cus-tomization since they get to purchase a product which is closerto their preferences at the same price that would be chargedif both firms were introduce a single product in no customiza-tion case. However, this is not the case for all customers withina firm’s product line. Those customers who are closer to thecorner than to the center of a firm’s product line accrue greatersurplus with customization than in the single product case. On


Fig. 6. Consumer Surplus Based on Location from the Center of a Firm’s Product.

the other hand, those customers who are closer to the center thanto the corner of a firm’s product line incur greater disutility withcustomization than in the single product case. The following re-sult summarizes the comparison of customization with no cus-tomization in terms of consumer surplus.

Corollary 5: 1) The average consumer surplus under targetedmass customization is higher than or equal to the average con-sumer surplus under mass customization. 2) However, the av-erage consumer surplus when any type of customization is of-fered is higher than the average consumer surplus when bothfirms offer only a single product.

Corollary 5 can be easily proven by referring to Fig. 6. As-sume that each firm locates its product variety which is theclosest to its competitor’s product line at distance away fromthe center of its own product line. A customer who is located

distance away from the center of a firm’s product line getsutility when firms customize, regard-

less of the type of the customization. Because is always 1/8under targeted mass customization and less than or equal to1/8 under mass customization, the average consumer surplusunder mass customization cannot be greater than that under tar-geted mass customization. This proves the first part of Corol-lary 4. On the other hand, the customer who is located dis-tance away from the center gets utility whenneither firm customizes. By comparing these two expressions,one can show that customers located at most distance awayfrom the center of a firm’s product line get higher surplus underno customization than under customization. Yet, customers lo-cated more than distance away from the center of a firm’sproduct line get higher surplus under customization than underno customization. Since customers are located uniformly, thetotal surplus gain with customization for customers who arelocated at where is equal to the total sur-plus loss with customization for customers who are located at

where . Hence, there is no change in the totalsurplus within the product line. However, customers located at

where get more surplus under customizationthan under no customization. Consequently, we can concludethat customization increases the average consumer surplus.

IV. CONCLUSION, DISCUSSIONS AND LIMITATITONS

Improvements in manufacturing technologies have made itpossible to customize almost every conceivable product. How-ever it is not clear how competing firms should approach cus-

tomization. In this paper we first analyze whether customiza-tion is a preferred choice when customization technologies arereadily available to every firm, or whether firms are better offwith no customization. Once we define conditions under whichcustomization is the optimal strategy when there is competition,we next address what kind of customization strategy- mass cus-tomization, targeted mass customization or product prolifera-tion- firms should adopt in a competitive setting.

Our analysis of customization strategies versus no customiza-tion has led to several interesting and new findings. Using amodel that compares no customization with mass customizationto address the first question, we find that mass customization isnot always the best strategy even if customization technology isvery cheap. We show that whether competing firms choose masscustomization over no customization depends upon firms’ costof obtaining customer preferences vis-à-vis the price premiumsassociated with customized products. No matter how cheap it isto invest in the customization technologies, firms choose to offera single product variety instead of mass customization when thecost of soliciting customer preference information is relativelyhigh compared to customers’ willingness to pay for customizedproducts. When firms adopt mass customization, the extent ofthe customization scope depends on the cost of customizationtechnology. If customization technology is sufficiently inexpen-sive, each firm offers customized products to 1/4 of the market.Otherwise, each firm’s customization scope covers less than 1/4of the market. A surprising result of our analysis is that even ifcustomization technology is inexpensive, not every customer isoffered a customized product. The reason is that extending cus-tomization beyond a certain limit triggers a hostile price war,which in turn hurts the competing firms. Another surprising re-sult is that the optimality of mass customization does not neces-sarily mean that mass customization generates more profit forfirms than no customization. A firm’s profit is greater undermass customization than under no customization only when thecost of customization is significantly small. When customiza-tion costs are high, the additional rent that firms can extract fromcustomers does not offset the cost of customization. This resultimplies that adoption of customization can result from competi-tion, not from customization’s positive effect on firms’ bottomlines. We also find that when customization gets cheaper, firms’profits may decline. The intuition behind this result is that firms,in such cases, aggressively extend their customization scopes,


thereby leading to a more severe competition, which in turn re-sults in a marginal return that is less than the marginal cost.

Given that customization can be the optimal strategy forcompeting firms, we address the second question using amore general model that allows firms to target customizationto specific customer points or segments. We show that masscustomization may not be the optimal customization strategyunless customization is costless (i.e., operational flexibility costand cost of eliciting customer information are zero). Insteadof mass customization, firms choose to adopt targeted masscustomization as a customization strategy and offer multiplecustomization scopes. Targeted mass customization also leadsto a larger product line than mass customization. Firms evenoffer multiple discrete products without any customizationsegment when the cost of learning customer preference struc-ture is expensive compared to price premiums resulting fromcustomization. Analogous to no customization versus masscustomization, we find that firms do not consider the cost ofoperational flexibility when choosing between targeted masscustomization and product proliferation.

Contrary to the expectation that customization reducesproduct differentiation and, therefore, leads to price warsamong competing firms, we show that neither mass customiza-tion nor targeted mass customization intensifies the pricecompetition. The reason for this counter-intuitive result is thatprice competition escalates cannibalization among a firm’sown product varieties. Each firm has more incentive to reducecannibalization than to extend its market reach through pricecompetition.

We show that targeted mass customization extracts largercustomer rent than mass customization. The reason is that, intargeted mass customization, each firm has an ability to placeconventional (noncustomization) segments into its product lineto prevent competition from the rival firm in those segments.However, in mass customization, conventional segments of afirm are open to direct competition from the rival firm. Wefinally show that customers are the primary beneficiaries ofany form of customization, because both mass customizationand targeted mass customization result in an average consumersurplus that is higher than the average consumer surplus in caseof no customization.

Our paper presents a theoretical framework to validate someobservations made by academics and practitioners alike. Forexample, Zipkin [58], when arguing about the limits of masscustomization, highlights that firms should “carefully assessthe technology and the market demand before committing” tomass customization. Our paper provides a systematic analysisof those operational and market factors to which Zipkin alludes.From a marketing and operations management perspective, ouranalysis suggests that firms may be better off concentratingon one or more niche regions within the product space whenadopting a customization strategy. For instance, Billington [6],in describing the customization strategy adopted by HewlettPackard for its printers, claims “you (firms) can mass customizeeffectively for only about 20% of the customers.” Our findingon the limit of customization can be interpreted as a moregeneral version of this early quantification.

A. Discussion and Managerial Implications

Customization is a reality in today’s manufacturing environ-ments. An important question within this context is to define anappropriate customization strategy. Our results provide usefulinsights to managers who are considering offering customizedproducts. We summarize our model implications for practicebelow.

Customization is not only a technological problem but also astrategic one. Firms carefully assess the cost of customizationagainst its benefits before investing in customization technolo-gies. Firms should know that they cannot customize to all cus-tomers. This is especially true in traditional goods markets. Ifgathering customer preferences is costly, firms should not evenconsider customization, no matter how cheap the technologythat enables it. However, Internet technologies have drasticallychanged the way firms learn customer preferences. Widespreadavailability of personalization services makes getting customerpreferences almost costless for firms. By creating personal pro-files, customers reveal their preferences to manufacturers. Cus-tomers also design their own products by interacting with on-line design sites. For example, Land’s End provides design-to-fitsolutions to address its customers’ tastes and needs. In addi-tion to customizing an item on several dimensions using a soft-ware-based product configurator, customers can make an itemuniquely theirs with monograms. So firms should try, as muchas possible, to use the technologies that permit customers toreveal their preferences directly rather than through traditionalmethods, such as market surveys. Once firms have systems inplace to get customer tastes, they should determine the levelof customization purely based on the cost of manufacturingtechnologies that provide flexibility for customization. If cus-tomization is not free, firms should target their customizationefforts only to some niche segments and serve other customersegments with standard products. An untargeted customizationstrategy cannot be the best strategy. Therefore, firms shoulddefine their niche segments before investing in customizationtechnologies. However, if price premiums associated with cus-tomized offerings are not large enough to cover the cost of cus-tomization, firms should consider offering just a few discreteproduct varieties.

Unlike traditional (nondigital) goods, once a digital productis ready, it is easy to modify it to meet the needs of various cus-tomer types. Firms do not have to spend extra money on cus-tomization technologies to produce customized digital goods.Further, firms can harness the power of the Internet to learn cus-tomer needs and deliver the customized digital content easily.Essentially, firms may not incur any cost for customization ofdigital goods on the Internet. For example, many news sites onthe Internet allow their readers to have personalized newslet-ters or personalized news pages. There is no need for targetedcustomization in this case (i.e., allowing only some readers tocustomize the appearance of the news pages). Therefore, firmsshould engage in mass customization for customization of dig-ital goods as long as eliciting customer preferences is almostcostless.


B. Limitations and Future Research Directions

Our paper has several limitations. Some of these limitationsare direct results of modeling assumptions. First, we assume thata firm’s product line consists of one or more segments. How-ever, optimal product line found in this paper cannot be directlymapped to product characteristics to operationalize customiza-tion choices since distribution of customer preferences can bedifferent for different products. Second, we assume that eachproduct variety within a product space costs the same. There-fore, the unit cost of production was normalized to zero in themodel. This is a reasonable assumption given that customizationtechnologies today enable production of varieties at the samemarginal cost. However, firms need to obtain customization ca-pabilities by investing in customization technologies to offercustomized product varieties. This cost obviously depends onthe extent of the customization scope, as captured in the model.Third, we assume that operational flexibility cost is a quadraticfunction of the customization scope. Quadratic cost functioncaptures the fact that marginal cost of customization flexibilityincreases as the scope of customization increases. Further, thiscost function has been used in prior studies [20]. However, anyfunctional form with increasing marginal cost, such as expo-nential cost function, is sufficient to get the same qualitative re-sults presented in this study. The issue of whether quadratic orexponential form captures the true nature of the customizationflexibility is an empirical one, and can be different for differentproducts. The functional form of operational flexibility cost mayonly affect the size of customization scope(s). Fourth, we as-sume that firms compete in a duopoly market. The analysis ofcustomization strategies when more than two firms compete inthe market can provide further insights into the extent of com-petition. Future research should address the above-mentionedlimitations of this paper.

Notwithstanding these potentially attractive avenues for fur-ther research, the present study, which we believe is the firstto investigate various customization strategies, provides usefulinsights into the value of customization under competition. Ourfindings justify the common belief that customization is not onlya technological problem, but also a strategic one. To benefit fromcustomization, firms have to address both of these problems.

APPENDIX

Proposition 1: The Lagrangian for the constrained opti-mization problem can be written as follows:

The Kuhn-Tucker conditions are

The Kuhn-Tucker conditions are satisfied in three regionswith different customization scopes.

a) When ,, .

b) When, .

c) When , ,.

These regions overlap only when andare satisfied. Therefore, in the overlapped

region, whichever customization alternative gives the higherpayoff is the optimum strategy. Because

, theoptimal strategy in the overlapped region is alternative c).

Proposition 2: The difference between each firm’s profitwhen both customize and that when neither customizes can beobtained with simple algebra from (12)

&

& .

When and , is always negative.However, when and , is positiveif .

Solution to Pricing Problem in Customization WithMultiple Scopes: After (16)–(18) are substituted in (15), thefollowing first-order conditions are obtained


Solving these equalities simultaneously leads to pricing givenin (19).

Proposition 3: The Lagrangian for the problem given in(14) such that (20)–(22) are satisfied is as follows.

By using the Lagrangian above, we can obtain the Kuhn-Tucker conditions as

Since firms are identical, the number of scopes should beidentical and equal to . The critical points

that satisfy the above K-T conditions.1) When , and .

From the second set of the K-T conditions, we can get. After substituting these

into the third set of the K-T conditions, we obtain the set ofequalities, shown in the equation at the bottom of the page,where is an by matrix whose all elements are 1,

is an by identity matrix, and is by 1 vectorwhose elements are ,

, ,,

, and . Then,the difference between the two rows of the matrix aboveimplies that ’s are equal. By solving fourth and fifthK-T conditions, we obtain . Therefore,

. Substituting into , we get. Also note that has to

be greater than zero to ensure feasibility.2) When , and .

From the second set of the K-T conditions, we can get. After substituting these into the third set of the

K-T conditions, we obtain the set of equalities, shown inthe equation at the top of the page, where

, , and. As in 1), the difference between the

two rows of the matrix above implies that ’s are equal.


By solving fourth and fifth K-T conditions, we obtain. Therefore, . From the first set

of K-T conditions, we get . To ensurethat it is nonnegative, we need to satisfy .Combining the results in 1) and 3) leads to Proposition 2.

REFERENCES

[1] D. M. Anderson, Agile Product Development for Mass Customiza-tion: How to Develop and Deliver Products for Mass Customiza-tion. Chicago: Niche Markets, JIT, Built-to-Order and FlexibleManufacturing, Irwin Professional Publishing, 1997.

[2] A. Ansari, N. Economides, and A. Ghosh, “Competitive positioning inmarkets with non-uniform preferences,” Marketing Sci., vol. 13, no. 3,pp. 248–273, 1994.

[3] G. Azzone and U. Bertele, “Comparing manufacturing systems withdifferent flexibility: A new approach,” in Proc. The Decision SciencesInstitute, Boston, MA, 1987, pp. 690–693.

[4] B. Bayus and W. Putsis, “Product proliferation: An empirical analysisof product line determinants and market outcomes,” Marketing Sci.,vol. 18, no. 2, pp. 137–153, 1999.

[5] M. Beckmann, “Spatial price policies revisited,” Bell J. Econ., vol. 7,pp. 619–630, 1976.

[6] J. Billington, “How to customize for the real world,” Harvard Manage.Update.

[7] J. Branden and J. Eaton, “Product line rivalry,” Am. Econ. Rev., vol. 74,no. 3, pp. 323–333, 1984.

[8] G. S. Carpenter, “Perceptual position and competitive brand strategy ina two-dimensional two-brand market,” Management Sci., vol. 35, no.9, pp. 1029–1044, 1989.

[9] H. Cavusoglu and S. Raghunathan, Product variety strategies undercompetition, Working Paper, University of Texas at Dallas, Richardson,TX, 2003.

[10] E. H. Chamberlin, The Theory of Monopolistic Competition, 8th ed.Cambridge, MA: Harvard Univ. Press, 1962.

[11] Y. Chen and G. Iyer, “Consumer addressability and customizedpricing,” Marketing Sci., vol. 21, no. 2, pp. 197–208, 2002.

[12] S. C. Choi, W. S. Desarbo, and P. T. Harker, “Product positioning underprice competition,” Management Sci., vol. 36, no. 2, pp. 175–199, 1990.

[13] J. Connor, “Food product proliferation,” Am. J. Agricultural Econ., vol.63, pp. 607–617, Nov. 1981.

[14] W. M. Cox and R. Alm, “The right stuff: America’s move to masscustomization,” in 1998 Annu. Rep.Federal Reserve Bank of Dallas,Dallas, TX, 1998.

[15] L. T. Cullen, “Have it Your Way,” Time Mag., Dec. 2002.[16] C. D’Aspremont and J. J. Gabszewicz, “On hotelling’s stability in com-

petition,” Econometrica, vol. 47, no. 5, pp. 1145–1150, Sep. 1979.[17] P. Desai, S. Kekre, S. Radhakrishnan, and K. Srinivasan, “Product dif-

ferentiation and commonality in design: Balancing revenue and costdrivers,” Management Sci., vol. 47, no. 1, pp. 37–51, 2001.

[18] P. S. Desai, “Quality segmentation in spatial markets: When does can-nibalization affect product line design?,” Marketing Sci., vol. 20, no. 3,pp. 265–283, 2001.

[19] R. Dewan, B. Jing, and A. Seidmann, “Adoption of internet-basedproduct customization and pricing strategies,” J. Manage. Inform.Syst., vol. 17, no. 2, pp. 9–28, 2000.

[20] ——, “Product customization and price competition on the internet,”Manage. Sci., vol. 49, no. 8, pp. 1055–1070, 2003.

[21] A. K. Dixit and J. E. Stiglitz, “Monopolistic competition and optimumproduct diversity,” Am. Econ. Rev., vol. 67, no. 3, pp. 297–308, Jun.1977.

[22] E. Feitzinger and H. Lee, “Mass customization at hewlett-packard: Thepower of postponement,” Harvard Bus. Rev., Jan. 1997.

[23] M. L. Fisher and C. D. Ittner, “The impact of product variety on auto-mobile assembly operations: Empirical evidence and simulation anal-ysis,” Manage. Sci., vol. 45, no. 6, pp. 771–786, 1999.

[24] M. L. Fisher, A. Jain, and J. P. MacDuffie, “Strategies for Product Va-riety: Lessons From the Auto Industry,” in Redesigning the Firm, B.Kogut and E. Bowman, Eds. New York: Oxford Univ. Press, 1995,pp. 116–154.

[25] D. Fudenberg and J. Tirole, Game Theory. Cambridge, MA: The MITPress, 1991.

[26] J. T. Gourville and J. J. Koehler, Downsizing price increases: A greatersensitivity to price than quantity in consumer markets, Working Paper,Harvard Business School, Cambridge, MA, 2004, pp. 04–042.

[27] P. Green and A. Krieger, “Models and heuristics for product line se-lection,” Marketing Sci., vol. 4, pp. 1–19, 1985.

[28] Y. P. Gupta and S. Goyal, “Flexibility of manufacturing systems: Con-cepts and measurements,” Eur. J. Operational Res., vol. 43, no. 2, pp.119–135, 1989.

[29] J. Hauser, “Competitive price and positioning strategies,” MarketingSci., vol. 7, no. 1, pp. 76–91, 1988.

[30] J. R. Hauser, “Metrics thermostat,” J. Product Innovation Manage., vol.18, no. 3, pp. 134–153, May 2001.

[31] T. Ho and C. Tang, Eds., Product Variety Management. Norwell,MA: Kluwer Academic, 1998.

[32] H. Hotelling, “Stability in competition,” Econ. J., vol. 39, no. 153, pp.41–57, Mar. 1929.

[33] A. Kaul and V. R. Rao, “Research for product positioning and designdecisions: An integrative review,” Int. J. Res. Marketing, vol. 12, pp.293–320, 1995.

[34] K. Kim and D. Chhajed, “Product design with multiple quality-typeattributes,” Manage. Sci., vol. 48, no. 11, pp. 1502–1511, 2002.

[35] V. Krishnan and K. T. Ulrich, “Product development decisions: A re-view of literature,” Manage. Sci., vol. 47, no. 1, pp. 1–21, 2001.

[36] K. Lancaster, “Socially optimal product differentiation,” Am. Econ.Rev., vol. 65, no. 4, pp. 567–585, Sep. 1975.

[37] ——, “The economics of product variety: A survey,” Marketing Sci.,vol. 9, pp. 189–210, 1990.

[38] H. L. Lee and C. S. Tang, “Modeling the costs and benefits of delayedproduct differentiation,” Manage. Sci., vol. 43, pp. 40–53, Jan. 1997.

[39] H. L. Lee, “Effective inventory and service management throughproduct and process redesign,” Operations Res., vol. 44, no. 1, pp.151–159, 1996.

[40] H. L. Lee and C. Billington, “Designing products and processes forpostponement,” in Management of Design: Engineering and Man-agement Perspectives, S. Dasu and C. Eastman, Eds. Boston, MA:Kluwer Academic, 1994, pp. 105–122.

[41] H. L. Lee and M. Sasser, “Product universality and design for supplychain management,” Production Planning Control (Special Issue onSupply Chain Management), vol. 6, no. 3, pp. 270–277, 1995.

[42] B. L. MacCarthy, P. G. Brabazon, and J. Bramham, “Key valueattributes in mass customization,” in Moving into Mass Customization:Information Systems and Management Principles, C. Rautenstrauch,R. Seelmann-Eggebert, and K. Turowski, Eds. Berlin, Germany:Springer-Verlag, 2002.

[43] J. P. MacDuffie, K. Sethuraman, and M. Fisher, “Product variety andmanufacturing performance: Evidence from the international automo-tive assembly plant study,” Manage. Sci., vol. 42, no. 3, pp. 350–369,1996.

[44] X. Martinez-Giralt and D. J. Neven, “Can price competition dominatemarket segmentation?,” J. Ind. Econ., vol. 36, no. 4, pp. 431–442, Jun.1988.

[45] K. S. Moorthy, “Product and price competition in a duopoly,” Mar-keting Sci., vol. 7, no. 2, pp. 141–169, 1988.

[46] M. Mussa and S. Rosen, “Monopoly and product quality,” J. Econ.Theory, vol. 18, pp. 301–317, Aug. 1978.

[47] S. Netessine and N. Rudi, “Centralized and competitive inventorymodels with demand substitution,” Operations Res., vol. 51, no. 2, pp.329–335, 2003.

[48] F. T. Piller, “Customize your time: mass customization in the watchindustry,” Mass Customization News, vol. 6, no. 1, May 2003.

[49] B. J. Pine, II, Mass Customization: The New Frontier in Business Com-petition. Boston, MA: Harvard Bus. School Press, 1993.

[50] W. Putsis and B. Bayus, “An empirical analysis of firms’ product linedecisions,” J. Marketing Res., vol. 38, pp. 110–118, Feb. 2001.

[51] T. Randall and K. Ulrich, “Product variety, supply chain structure, andfirm performance: Analysis of the U.S. bicycle industry,” Manage. Sci.,vol. 47, pp. 1588–1604, 2001.

[52] S. C. Salop, “Monopolistic competition with outside goods,” Bell J.Econ., vol. 10, no. 1, pp. 141–156, 1979.

[53] S. Smith and N. Agrawal, “Management of multi-item retailer inven-tory systems with demand substitution,” Operational Res., vol. 48, pp.50–64, 2000.

[54] J. F. Thisse and X. Vives, “On the strategic choice of spatial pricepolicy,” Am. Econ. Rev., vol. 78, pp. 122–137, 1988.

[55] M. M. Tseng and J. Jiao, “Mass customization,” in Handbook of Indus-trial Engineering, G. Salvendy, Ed., 3rd ed. New York: Wiley, 2001,pp. 684–709.

[56] R. K. Tyagi, “Sequential product positioning under differential costs,”Manage. Sci., vol. 46, no. 7, pp. 928–940, 2000.


[57] M. L. Weitzman, “Monopolistic competition with endogenous special-ization,” Rev. Econ. Studies, vol. 61, pp. 45–56, 1994.

[58] P. Zipkin, “The limits of mass customization,” MIT Sloan Manage.Rev., vol. 42, no. 3, pp. 81–87, 2001.

Hasan Cavusoglu received the Ph.D. degree inmanagement science with a specialization in Man-agement Information Systems (MIS) from theUniversity of Texas at Dallas, Richardson, in 2003.

He is currently an Assistant Professor of MIS atthe Sauder School of Business, University of BritishColumbia, Vancouver, BC, Canada. His currentresearch interests are the economics of informationsystems, economics of information security, andmanagement of technology and information. He haspublished in Communications of the AIS.

Dr. Cavusoglu is a member of INFORMS.

Huseyin Cavusoglu received the Ph.D. degreein management science with a specialization inmanagement information sciences (MIS) from theUniversity of Texas at Dallas, Richardson, in 2003.

He is currently an Assistant Professor of Man-agement Information Systems in the School ofManagement at the University of Texas at Dallas,Richardson. He has published in InformationSystems Research, Communications of the ACM,INFORMS Decision Analysis, International Journalof Electronic Commerce, and Communications of

the AIS. His major research interests are in the areas of information economics,assessment of the value of IT security, and IT security management.

Dr. Cavusoglu is a member of AIS and INFORMS.

Srinivasan Raghunathan received the Ph.D. de-gree in business from the University of Pittsburgh,Pittsburgh, PA, in 1990.

He is currently an Associate Professor ofManagement Information Systems in the Schoolof Management at the University of Texas atDallas, Richardson. He has published in Man-agement Science, Information Systems Research,Journal of Management Information Systems,and IEEE TRANSACTIONS ON KNOWLWDGE AND

DATA ENGINEERING and IEEE TRANSACTIONS

ON SYSTEMS, MAN, AND CYBERNETICS. His research interests concern theeconomics of information systems.

12 ieee transactions on engineering management, vol. 54, no. 1

Documents